Fig. 4.

Non-maximum suppression: we utilise the Matlab function ‘findpeaks’ as a reference implementation for non-maximum suppression. That is, from a local maximum on $D(u,\mathit{\sigma })$, the horizontal distance to $D(u,\mathit{\sigma })$ is measured to define left and right intervals ${\text{min}}_{\mathrm{L}}={\text{min}}_{{\mathrm{interval}}_L}(D(u,\mathit{\sigma })),{\text{min}}_{\mathrm{R}}$ is defined likewise. Subsequently, max(${\text{min}}_{\mathrm{L}},{\text{min}}_{\mathrm{R}}$) is taken as a reference level. The prominence of each local maximum is then computed as the difference between the value of $D(u,\mathit{\sigma })$ at the local maximum and the reference level. Low prominence peaks are suppressed. If either interval reaches the end of the signal, we set its minimum to be zero